Neil Chowdhury

I'm a Member of Technical Staff at Thinking Machines Lab.

I work on AI safety and alignment. Previously, I was a founding engineer at Transluce and a safety researcher at OpenAI.

Publications

Introducing SWE-bench Verified
A human-validated benchmark for more reliable evaluation of AI coding capabilities.
Neil Chowdhury*, James Aung*, Chan Jun Shern*, Oliver Jaffe*, Dane Sherburn*, Giulio Starace*, Evan Mays*, Rachel Dias, Marwan Aljubeh, Mia Glaese, Carlos E. Jimenez, John Yang, Kevin Liu, Aleksander Madry
Surfacing Pathological Behaviors in Language Models
A new red-teaming objective for discovering rare, harmful behaviors in language models.
Neil Chowdhury, Sarah Schwettmann, Jacob Steinhardt, Daniel D. Johnson
- Blog (2025)
- X Thread
MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering
Testing AI agents on 75 Kaggle machine learning competitions.
Jun Shern Chan*, Neil Chowdhury*, Oliver Jaffe*, James Aung*, Dane Sherburn*, Evan Mays*, Giulio Starace*, Kevin Liu, Leon Maksin, Tejal Patwardhan, Lilian Weng, Aleksander Madry
- ICLR 2025 (Oral)
- arXiv
- GitHub
WeirdChat: A catalog of unexpected AI behaviors, discovered automatically
A public catalog of unexpected behaviors in frontier open-weight models with 175,000+ annotated transcripts.
Neil Chowdhury, Cassidy Laidlaw, Kaiying Hou, Daniel Johnson, Sarah Schwettmann, Jacob Steinhardt
- Blog (2026)
- Dataset